Howdy.

Jesta' Pizza Mobile App

Video Demo

A Swift-based mobile app tailored for UT Austin students to customize, order, and track their pizzas from Jesta’ Pizza in real-time. Additionally, there is a dedicated staff section to manage those orders and their respective status. Built with Core Data for offline data persistence and Firebase for seamless authentication and user management.

Implements SwiftUI components, such as UITableView for order history and favorites, for detailed pizza descriptions, and UISegmentedControl to manage order statuses. A UIScrollView enhances navigation, while a custom-built calendar view allows pickup scheduling.

Backend integration utilizes Firebase SDK to handle login, registration, and secure user sessions. Push notifications provide real-time updates on order statuses, while Core Data enables caching and offline functionality for smooth user experiences even without connectivity.

Scrapalot: The Go-To Scrapbooking Platform

Visit Scrapalot

(Ongoing learning experience) The go-to scrapbooking platform that allows you to upload scraps to share with others and then place them into your own custom scrapbooks for inspiration and creativity.

This is a full-stack web application relying on Next.js 14 with TypeScript, incorporating both server and client components to balance performance and interactivity.

On the backend, Supabase handles secure authentication, real-time database capabilities, and efficient data management for holding scrap data, user information, etc. For context, custom-made context providers were made (e.g., UserContext, ScrapContext) for efficient global state management. Furthermore, there is implementation of React patterns such as higher-order components and custom hooks to handle when a user should be redirected to the login page (if not logged in). The user interface is built with Tailwind CSS to offer a responsive and dynamic experience.

Additionally, important parts of the application such as form handling (with react-hook-form), comprehensive error handling, and performance optimization techniques such as lazy loading and code splitting have been accounted for. Security best practices were applied (e.g., protected routes and secure environment variable management).

Ultimately, throughout the development process, I have applied my learning in API development, version control with Git, and preparation for production deployment, resulting in a polished, high-performance web application ready for real-world use.

Comprehensive Regression Model Performance Evaluation

Evaluates multiple regression models including Lasso, Ridge, Decision Tree, Random Forest, and Neural Network on a .csv dataset containing stock returns using Python (libraries: pandas, numpy, sklearn (linear_model and metrics)). Implements shrinkage techniques for Lasso and Ridge, compares their MSEs, and analyzes model performance. Trains and predicts using Decision Tree, Random Forest, and Neural Network, assessing their MSEs and model depths to determine the most effective approach.

Texas InventionWorks 3D Print Job Software Manager

Manages 3D print jobs within a budget, optimizing for high-quality prints using queues and hash tables using Python. Allows selection of detail and material. Utilizes the knapsack algorithm to maximize output, reviewing and simulating jobs based on budget constraints.

DeFi Protocol Cost of Equity Derivation

Downloads 5 years of stock data for SPY and AAVE, computes daily returns, and renames columns using Python (libraries: yfinance, pandas, statsmodels). Utilizes advanced regression analysis to derive AAVE's beta relative to SPY. Applies the CAPM formula to precisely calculate the cost of equity, enhancing investment decision-making for DeFi protocols.

Web Swinging Simulation with Dynamic Physics and Momentum

Engineered a Lua-based web swinging simulation, featuring dynamic rope creation and in-engine physics to enable realistic swinging between buildings. Utilized custom attachments and force application logic, including vector force momentum and rope constraints, ensuring accurate attachment to the player's torso. Comprehensive checks for valid targets, distance constraints, and custom animations.

Mass E-Mail Automation Application Using Gmail API

Developed a JavaScript application leveraging the Gmail API to draft and schedule customizable emails, automating the process of sending 125+ marketing emails weekly. Saved over 20 hours per week and achieved a follow-on conversion rate of ~11%. Enhanced lead conversion performance by 10% through tailored customizability options specific to various industries.